Predicting the EQ-5D-5L Utility Scores From the Impact of Vision Impairment Questionnaire in Thai Patients

Abstract

Objectives: This study aims to predict the EQ-5D-5L utility scores from the impact of vision impairment (IVI) questionnaire in Thai patients using mapping techniques. Methods: This is a secondary data analysis. A total of 499 patients with multiple levels of visual impairment were recruited from King Chulalongkorn Memorial Hospital in Thailand between February and July 2022. Ordinary least square, Tobit, censored least absolute deviation, and adjusted limited dependent variable mixture model regression models were used to map the IVI questionnaire onto EQ-5D-5L index scores. IVI domain scores, IVI total score, gender, age, employment status, and best corrected visual acuity were included as predictors. Performance metrics including root mean square error, mean absolute error, and adjusted R² were used to determine the best predictive model. Results: The results indicated that EQ-5D-5L index scores were significantly associated with the reading and emotional well-being domains of the IVI. Among sociodemographic and clinical variables, higher age score was significantly associated with lower EQ-5D index scores (P < .01). The mean predicted EQ-5D-5L value (0.803) was lower than the mean observed value (0.808). The adjusted limited dependent variable mixture model 1-component model demonstrated the best predictive performance (root mean square error 0.137, mean absolute error 0.101, adjusted R² 0.689). Conclusions: Mapping techniques effectively predicted EQ-5D-5L utility scores from the IVI questionnaire in Thai patients. The predicted model enhances decision analysis by capturing health utility values, informing quality-adjusted life-years, and supporting health economic evaluations when vision-specific measures are available.